Continuous Target Shift Adaptation in Supervised Learning

Tuan Duong Nguyen, Marthinus Christoffel, Masashi Sugiyama
Asian Conference on Machine Learning, PMLR 45:285-300, 2016.

Abstract

Supervised learning in machine learning concerns inferring an underlying relation between covariate \bx and target y based on training covariate-target data. It is traditionally assumed that training data and test data, on which the generalization performance of a learning algorithm is measured, follow the same probability distribution. However, this standard assumption is often violated in many real-world applications such as computer vision, natural language processing, robot control, or survey design, due to intrinsic non-stationarity of the environment or inevitable sample selection bias. This situation is called \emphdataset shift and has attracted a great deal of attention recently. In the paper, we consider supervised learning problems under the \emphtarget shift scenario, where the target marginal distribution p(y) changes between the training and testing phases, while the target-conditioned covariate distribution p(\bx|y) remains unchanged. Although various methods for mitigating target shift in classification (a.k.a. \emphclass prior change) have been developed so far, few methods can be applied to continuous targets. In this paper, we propose methods for continuous target shift adaptation in regression and conditional density estimation. More specifically, our contribution is a novel importance weight estimator for continuous targets. Through experiments, the usefulness of the proposed method is demonstrated.

Cite this Paper


BibTeX
@InProceedings{pmlr-v45-Nguyen15, title = {Continuous Target Shift Adaptation in Supervised Learning}, author = {Nguyen, Tuan Duong and Christoffel, Marthinus and Sugiyama, Masashi}, booktitle = {Asian Conference on Machine Learning}, pages = {285--300}, year = {2016}, editor = {Holmes, Geoffrey and Liu, Tie-Yan}, volume = {45}, series = {Proceedings of Machine Learning Research}, address = {Hong Kong}, month = {20--22 Nov}, publisher = {PMLR}, pdf = {http://proceedings.mlr.press/v45/Nguyen15.pdf}, url = {https://proceedings.mlr.press/v45/Nguyen15.html}, abstract = {Supervised learning in machine learning concerns inferring an underlying relation between covariate \bx and target y based on training covariate-target data. It is traditionally assumed that training data and test data, on which the generalization performance of a learning algorithm is measured, follow the same probability distribution. However, this standard assumption is often violated in many real-world applications such as computer vision, natural language processing, robot control, or survey design, due to intrinsic non-stationarity of the environment or inevitable sample selection bias. This situation is called \emphdataset shift and has attracted a great deal of attention recently. In the paper, we consider supervised learning problems under the \emphtarget shift scenario, where the target marginal distribution p(y) changes between the training and testing phases, while the target-conditioned covariate distribution p(\bx|y) remains unchanged. Although various methods for mitigating target shift in classification (a.k.a. \emphclass prior change) have been developed so far, few methods can be applied to continuous targets. In this paper, we propose methods for continuous target shift adaptation in regression and conditional density estimation. More specifically, our contribution is a novel importance weight estimator for continuous targets. Through experiments, the usefulness of the proposed method is demonstrated.} }
Endnote
%0 Conference Paper %T Continuous Target Shift Adaptation in Supervised Learning %A Tuan Duong Nguyen %A Marthinus Christoffel %A Masashi Sugiyama %B Asian Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2016 %E Geoffrey Holmes %E Tie-Yan Liu %F pmlr-v45-Nguyen15 %I PMLR %P 285--300 %U https://proceedings.mlr.press/v45/Nguyen15.html %V 45 %X Supervised learning in machine learning concerns inferring an underlying relation between covariate \bx and target y based on training covariate-target data. It is traditionally assumed that training data and test data, on which the generalization performance of a learning algorithm is measured, follow the same probability distribution. However, this standard assumption is often violated in many real-world applications such as computer vision, natural language processing, robot control, or survey design, due to intrinsic non-stationarity of the environment or inevitable sample selection bias. This situation is called \emphdataset shift and has attracted a great deal of attention recently. In the paper, we consider supervised learning problems under the \emphtarget shift scenario, where the target marginal distribution p(y) changes between the training and testing phases, while the target-conditioned covariate distribution p(\bx|y) remains unchanged. Although various methods for mitigating target shift in classification (a.k.a. \emphclass prior change) have been developed so far, few methods can be applied to continuous targets. In this paper, we propose methods for continuous target shift adaptation in regression and conditional density estimation. More specifically, our contribution is a novel importance weight estimator for continuous targets. Through experiments, the usefulness of the proposed method is demonstrated.
RIS
TY - CPAPER TI - Continuous Target Shift Adaptation in Supervised Learning AU - Tuan Duong Nguyen AU - Marthinus Christoffel AU - Masashi Sugiyama BT - Asian Conference on Machine Learning DA - 2016/02/25 ED - Geoffrey Holmes ED - Tie-Yan Liu ID - pmlr-v45-Nguyen15 PB - PMLR DP - Proceedings of Machine Learning Research VL - 45 SP - 285 EP - 300 L1 - http://proceedings.mlr.press/v45/Nguyen15.pdf UR - https://proceedings.mlr.press/v45/Nguyen15.html AB - Supervised learning in machine learning concerns inferring an underlying relation between covariate \bx and target y based on training covariate-target data. It is traditionally assumed that training data and test data, on which the generalization performance of a learning algorithm is measured, follow the same probability distribution. However, this standard assumption is often violated in many real-world applications such as computer vision, natural language processing, robot control, or survey design, due to intrinsic non-stationarity of the environment or inevitable sample selection bias. This situation is called \emphdataset shift and has attracted a great deal of attention recently. In the paper, we consider supervised learning problems under the \emphtarget shift scenario, where the target marginal distribution p(y) changes between the training and testing phases, while the target-conditioned covariate distribution p(\bx|y) remains unchanged. Although various methods for mitigating target shift in classification (a.k.a. \emphclass prior change) have been developed so far, few methods can be applied to continuous targets. In this paper, we propose methods for continuous target shift adaptation in regression and conditional density estimation. More specifically, our contribution is a novel importance weight estimator for continuous targets. Through experiments, the usefulness of the proposed method is demonstrated. ER -
APA
Nguyen, T.D., Christoffel, M. & Sugiyama, M.. (2016). Continuous Target Shift Adaptation in Supervised Learning. Asian Conference on Machine Learning, in Proceedings of Machine Learning Research 45:285-300 Available from https://proceedings.mlr.press/v45/Nguyen15.html.

Related Material